Numerous studies have found that emotionally arousing faces or scenes capture visual processing resources. Here we investigated whether emotional distractor words capture attention in an analogous way. Participants detected brief intervals of coherent motion in an array of otherwise randomly moving squares superimposed on words of positive, neutral or negative valence. Processing of the foreground task was assessed by behavioural responses and steady-state visual evoked potentials (SSVEPs) elicited by the squares flickering at 15 Hz. Although words were task-irrelevant, P2 and N400 deflections to negative words were enhanced, indicating that emotionally negative word content modulated lexico-semantic processing and that emotional significance was detected. In contrast, the time course of behavioural data and SSVEP amplitudes revealed no interference with the task regardless of the emotional connotation of distractor words. This dissociation of emotion effects on early perceptual versus lexical stages of processing suggests that written emotional words do not inevitably lead to attentional modulation in early visual areas. Prior studies have shown a distraction effect of emotional pictures on a similar task. Thus, our results indicate the specificity of emotion effects on sensory processing and semantic encoding dependent on the information channel that emotional significance is derived from.